Machine learning to predict the product of chemical reactions using CORELS (Certifiably Optimal RulE ListS), by representing chemical species as bit vectors of fixed size and training one model for each bit of the representation of the product species.